Mild language production difficulties, particularly in confrontation naming, have been documented previously in patients with left temporal lobe epilepsy (LTLE). These difficulties, however, do not seem to be reflected in the conversational speech of these patients. In order to compare speech fluency in patients with left and right temporal lobe epilepsy (RTLE) experimentally, we studied global pause-to-speech ratios in patients with unilateral temporal lobe epilepsy, using cases with primary generalized epilepsy (PGE) as controls. We hypothesized that left-sides cases would exhibit longer pause durations than right-sided cases. Speech samples were elicited by asking subjects to describe five different New Yorker cartoons, with three repetitions of each description. Phonation was converted to graphic output, and pauses between 200 ms and 4000 ms were summated and divided by total phonation time. This measure did not discriminate significantly between the groups, although the LTLE group tended to pause longer than the RTLE or PGE groups. Increased variability in pause duration in the LTLE group during cycle 1 suggested that some individuals with LTLE are vulnerable to disruption when planning demands are high. A post hoc correlational analysis showed that variation in fluency was primarily explained by orthographically-based lexical retrieval, suggesting that individual differences in fluency are related to limitations in a high-level capacity relevant to the production of speech. It is unlikely that such limitations are specific to LTLE.